| Abstract |
Over the past decade, a new seismic imaging technique has emerged. Known as "ambient noise correlation", the technique allows us to perform tomographic imaging without deterministic sources. The cross-correlation function (CCF) of long noise records has been proved to converge toward the Green’s function between each pair of stations (Lobkis & Weaver, 2001; Shapiro & Campillo, 2004; Sabra et al., 2005a). Therefore, tomographic imaging can be applied using all possible pairs of stations over a network (Shapiro et al. 2005; Sabra et al. 2005b). The resolution of the recovered seismic velocity models only depends on the number of stations and the geometry of the network. Beyond its use in seismic tomography, the continuous nature of seismic noise can also be exploited to observe subtle variations in the seismic velocity or the diffracting character of the crust (Wegler & Sens-Schönfelder, 2006; Brenguier et al., 2008). Using a very dense seismic network designed to observe the first 5 km of the crust around the two geothermal sites of Soultz-sous-forêts and Rittershoffen, we compute the CCFs over time periods ranging from a few months to several years. Taking the characteristics of the noise recorded between 0.2s and 5s into account, we investigate the reliability of the Green’s function reconstruction as well as the ability to monitor speed variations induced at depth by geothermal activities. |